Automatic gain control apparatus and method

20~1982 AUTOMATIC GAIN CONTROL APPARATUS Ai~D METHOD Field of the Invention The present invention relates generally to communications receivers, and, more particularly, to a digital communication receiver with an automatic gain control apparatus and method which adjusts the gain of an amplifier using a control signal determined by the selective average energy level of a present input signal, the transfer function characteristic of the amplifier and the previous control signal.

Reference is made to U.S.

Patent 5,150,384 issued on September 22, 1992, filed on - September 28, 1990, "A Carrier Recovery Method and Apparatus Having an Adjustable Response Time Determined by Carrier Signal Parametersn, filed on behalf of Stephen Vincent Cahill on the same date herewith, containing related subject matter, and assigned to the assignee of the present invention.

Background of the Invention A radio communications system is comprised, at minimum, of a transmitter and a receiver.

The transmitter and the receiver are interconnected by a radio-frequency channel to permit transmission of an information signal therebetween. A digital receiver will generally include an amplifier with a gain adjusted by a controi signal.

The process of adjusting the gain of a received signal using a control signal is called Automatic Gain Control (AGC).

Although the concept of AGC in radio signal reception is well understood, automatic gain control of Time Division Multiple Access (TDMA) signals presents new challenges to the land-mobile industry.

Increased usage of cellular communications systems has resulted, in many instances, in the full utilization of 2051982 every available transmission channel of the frequency band allocated for cellular radio telephone communication.

In wide band TDMA systems, such as the cellular system proposed for use in the United States, hereinafter called 5 USDC for United States Digital Cellular, an RF channel is shared (time-division-multiplexed) among numerous subscribers attempting to access the radio system in certain ones of various time-division-multiplexed time - slots.

This permits transmission of more than one signal 10 at the same frequency, using the sequential time-sharing of a single channel by several radio telephones.

The time slots are arranged into periodically repeating frames thus, a radio communication of interest may be periodically discontinuous wherein unrelated signals are interleaved 15 with signals transmitted in other time slots.

The unrelated signals (of widely varying strength) must not influence the gain control of the signals of interest.

Varying signal strength may be caused by distance loss or multipath fading. A formidable challenge then is to 2 0 provide Automatic Gain Control of these periodically discontinuous TDMA signals.

The challenge is further enhanced by attempting to provide digital AGC in inexpensive receivers, specifically, those having Analog-to-Digital converters tA/D) with 25 limited dynamic range.

Since the received signal strength may vary by as much as 120dB in the land-mobile environment, but 8-bit Analog-to-Digital (A/D) converters, moderately priced, for digital signal processing are limited to 48dB dynamic range, techniques must be 30 developed for controlling the gain of the portions of the radio receiver prior to the A/D converters to keep the signal at the input to the A/D converters within the limited dynamic range of the A/D converters.

The challenge then is to handle a 120dB range discontinuous 3 5 signal with a 48dB device; otherwise, prohibitively 20~1982 expensive A/Ds with greater dynamic range must be uti I ized .

Certain operations necessary for new communications system protocols require the subscriber 5 unit to tune its receiver to another channel, measure signal strength, and report the measured signal strength to the system.

This operation provides the system with information about which channel has the strongest signal and hence can offer the best level of service to the 10 subscriber unit.

The system may then direct the unit to change the channel with which it communicates with the system.

This process is called Mobile Assisted Handoff, or MAHO.

This operation provides a very limited time for 15 measuring the power of the to-be-tested channel, since the testing operation must be done during idle periods of the discontinuous signalling.

Therefore, a fast AGC response time is needed to normalize the received signal strength energy level within the range of the A/D 2 0 converters.

Conventional AGC control loops are difficult to implement with rapid and accurate response times.

This is because AGC control loops, conventionally implemented, include the channel selectivity filters or matched filters in the contl~ol loop.

These filters have multiple poles and 25 possibly zeros, hence the gain control signal must be filtered additionally by a low-pass filter with a much lower cutoff frequency to insure that the loop response is stable.

This low-pass filter, for a conventional AGC system, has a slow response time, and hence limits the 30 response time of the AGC loop.

Additional A/D converter and signal processing methods have delay which also adds to the loop's response time and instability.

Conventional operation for AGC function using a signal processor requires a periodic calculation cycle, at the end 35 of which a new gain control signal is generated which 20S198~ replaces the present gain control signal.

This periodic calculation cycle is an artifact of the discrete-time nature of signal processors.

Since the received signal energy level is measured after the receiver's base band filtering and the A/D converters, the average signal energy level measured over the span of time since the last calculation cycle was performed and the last gain control signal was generated, will include both energy level measurements generated by the receiver response to the present gain control signal, as well as energy level measurements generated by the receiver response to the previous gain control signal, Thus, when a new gain control signal is issued by the signal processing element, a period of time elapses before the signals present at the inputs to the signal processing element begin to respond to those changes, and a further period of time elapses before the response to the new gain control signal has completed.

The energy level effects remaining from the previous 2 0 received signal, received during present calculation cycle span, but in response to the previous control signal, result in an inaccurate estimate of the present received signal energy level, and hence potential instability in the AGC control loop.

The AGC problem is further complicated by the fact that the input signal level multiplies the gain of the control loop, hence, when a gain error due to change in received signal level occurs, the loop response time is affected by how much gain error is present.

Yet another challenge for automatic gain control is introduced by variation in gain of the receiver signal amplifiers, and in effective delay through receiver filters, due to temperature and manufacturing tolerances.

For a conventional AGC method, further restriction of the loop's 2051982 response time must be made to assure stability with variations in gain and filter characteristics.

This invention takes as its initiative to overcome these challenges and realize certain advantages, presented 5 below.

Summary of the Invention An automatic gain control apparatus includes an 10 adjustable gain amplifier and a signal delay means.

The adjustable gain amplifier has a gain, re~ating an amplified output signal to a received input signal having time varying energy levels responsive to a control input signal, and having a transfer function with an associated inverse 15 transfer function relating the gain to the control input signal.

The amplified output signal and the control input signal incur time delays determined by the signal delay means.

An energy level error correction signal is determined responsive to a present amplified output signal 20 and a present control input signal. A previous control input signal is determined responsive to the signal delay means.

A next control input signal is determined responsive to the energy level error correction signal and the previous control input signal to generate a next amplified output 25 signal.

Brief Description of the Drawings FIG. 1 is a block diagram of a transmitter and 30 receiver which may employ the present invention.

FIG. 2 is a block diagram of a signal processor which may employ the present invention.

FIG. 3 is a signal timing diagram showing a delay caused by the loop's slow response time which may employ 3 5 the present invention.

20S1982 FIG. 4 is a block diagram of an automatic gain controller which may employ the present invention.

FIG. 5 is a block diagram of a control signal sensitivity estimator which may employ the present 5 invention.

Detailed Description of the Preferred Embodiment FIG. 1 shows a transmitter 101 and a receiver 103 10 which may employ the present invention.

The simplified block diagram of a receiver 103 illustrates automatic gain control in a digital quadrature receiver.

In general, the receiver 103 illustrates a quadrature demodulator having In-phase and Quadrature phases (I/Q) 107, Analog-to15 Digital converters (A/D)1 09, a signal processor 1 1 1 , andDigital-to-Analog converters (D/A) 1 17, 1 13, 121 providing control signals to a voltage controlled oscillator 119, a speaker 115, and an adjustable gain amplifier 105.

In the receiver operation, the signal is received, 2 0 amplified by an adjustable gain amplifier 105, quadrature demodulated (I/Q) into In-phase and Quadrature components using quadrature demodulator circuitry 107;

digitized by Analog-to-Digital converters of limited dynamic range (A/D) 109; and converted to a power sample 25 in signal processor 111 to detect the energy level of the signal.

The signal processor 111 has three outputs.

The first output goes to the Digital-to-Analog converter (D/A) 113 which is coupled to the speaker 115.

The speaker 115 produces a voice signal which is an estimate of the voice 3 0 signal sent from the transmitter 101, said voice signal being a portion of the information present in the received signal, quadrature encoded prior to transmission by transmitter 101, and decoded in signal processor 1 11 using known techniques.

The second output of the signal 35 processor 111 is coupled to the D/A converter 117.

The 20~1982 output of the D/A converter 117 is a phase error estimate signal used to adjust the phase of the voltage controlled oscillator 1 19.

The voltage controlled oscillator 1 19 produces a reference phase for the quadrature demodulator 107.

Reference is made to U.S.

Patent 5,150,384 , "A Carrier Recovery Method and Apparatus Having an Adjustable Response Time Determined by Carrier Signal Parametersn, filed on behalf of Stephen Vincent Cahill on the same date herewith, containing related subject matter and assigned to the assignee of the present invention.

The third output of the signal processor 111 is coupled to a D/A converter 121.

The output of the D/A converter 121 is control signal coupled to the adjustable gain amplifier 105.

The received signal energy level may vary due to multipath fading conditions, and varition in distance from the transmitter.

The digitized quadrature signals (I /Q') may have an energy level which is too high or too low when compared to an expected normalized value.

The preferred embodiment of the present invention discloses an apparatus and method which generates a control signal as an input to the adjustable gain amplifier 105 such that the change in received signal energy level at the output of the adjustable gain amplifier 105 is minimized, and the generation of said control signal is performed using an apparatus and method which results in a response to changes in received signal level which produces substantially correct signal level at the input to the A~D converters after one signal processing calculation cycle.

As the received signal energy level changes, a new control signal is derived for the input of the adjustable gain amplifier 105 using the estimated gain-versuscontrol signal transfer function of the adjustable gain amplifier 105. A change in energy level of the received signal from the output of the adjustable gain amplifier 2051982 105 is processed in the signal processor 111 to generate a corrected control signal.

The corrected control signal adjusts the gain of the amplifier 105 which produces an energy level output from the adjustable gain amplifier 105 5 with zero error.

The preferred embodiment of the present invention provides the advantage over other types of AGC methods in that it rapidly adjusts the energy level input to the signal processor 111 to a normalized value by means of selectively averaging the signal energy presented to the 10 signal processor, discarding those portions of the signal energy which accumulate during the early period of time in the signal processing calculation cycle when the effects of the previous gain control signal are still present.

This discarding of delayed signal energy permits the generation 15 of a gain control signal which exactly corrects the energy level to a normalized value, by means of the determination of the error in present gain, unaffected by previous gain control signal effects.

Now referring to FIG. 2 there is shown a block 20 diagram of signal processor 111 which may employ the present invention.

Digitized quadrature input signals l /Q are coupled to a rotator 201.

The rotator 201 has an oscillator which is controlled by a phase error estimate signal 203.

The rotator 201 produces quadrature outputs, 25 I'' and Q, having zero phase error and unity gain.

The information decoder 205, receives its input signals, I and Q'', to produce an approximation to the originally transmitted signal.

The carrier recovery function uses its received 30 quadrature inputs, I and Q'', to produce a phase error estimate signal 203 to adjust the phase of a reference signal relative to the phase of a received signal.

The purpose of a gain controller 209 is to generate a control signal (VCTL) at its output determined by its 35 quadrature input signals 1 and Q .

The control signal is 2051982 used to adjust the gain of the adjustable gain amplifier 105.

The adjustable gain amplifier 105 normalizes the received signal energy levels prior to the A/D converters 109 to fall within the useful dynamic range of the A/D 5 converters 109.

Now referring to FIG. 3 there is shown a timing diagram of the delay caused by the loop's response time prior to a signal processor 111.

The timing diagram shows three signal patterns; a control signal 301, an energy level 10 at the signal processor 303, and selective averaging of a signal 305.

The control signal 301 is the adjustable control input to the adjustable gain amplifier 105.

The control signal 301 is divided into three frames.

The first frame is the time slot of the previous signal 307.

The 15 second frame is the time slot of the present signal 309.

The third frame is a time slot for the next signal 311.

The signal state of the energy level at the signal processor 303 is also divided into the same three time frames.

The energy level of the signal at the processor 303 is offset, 20 or delayed, from the control signal 301 due to the slow response time of the matched filters in the quadrature demodulator 107 and the A/D converters 109.

One time slot for the present signal 309 is included between the event markers 313 and 315.

During that time slot, the 25 energy level at the signal processor 303 actually measures a portion of the previous signal 317 and a portion of the present signal 319.

It is not desirable to include information from the previous signal 317 when determining an average energy level for the present signal 30 319.

The preferred embodiment of the present invention discloses the advantage of selective averaging shown by the selective averaging signal 305.

The selective averaging signal 305 shows good and bad information in each time frame.

For any one time frame we only wish to 3 5 keep the signal information from the present signal which 20~1982 1 0 is good information 321 and discard energy level measurements from the previous signal which are bad 323.

The selective process of only averaging energy level measurements from the present signal 321 and not previous signals 323 is called selective averaging.

Now referring to FIG. 4 there is shown a block diagram of an automatic gain controller which may employ the present invention.

Phased corrected quadrature component signals 1 ' and Q are coupled to selective averager 401.

The selective averager 401 produces a measure of the average energy level of the present received signal 321 while discarding information from the previously received signal 323.

The output of the selective averager 401 may be described with the following equation:

Eout = (I/M)[~;(frm N-M to N) (l~ 2+Q 2)] where Eout is the average energy level output of selective averager 401, M is the amount of delay through the recever filters and A/D devices, measured in A/D samples, N is the number of samples per calculation cycle of the AGC operation, and l and Q are inputs to selective averager 401 The average energy measurement of the present signal 321 is coupled to a feedback gain adjuster 403 to produce a scaled normalized output.

The feedback gain adjustor 403 scales the energy level to a useful level for further processing in a transfer function inversor 405.

The scaling factor is a constant determined by hardware in the receiver 103.

Transfer function inversor 405 produces an energy level error correction signal output using inputs of the average scaled energy value from the feedback gain adjustor 403 and a slope estimate of gain versus control 20S1982 1 1 signal of the receiver adjustable-gain amplier 105 from the control signal sensitivity estimator 407.

The transfer function inversor 405 includes an inverse gain control transfer function estimate of the adjustable gain 5 amplifier 105.

Usually the gain control transfer function of an adjustable gain amplifier 105 is exponential times a constant.

In this case, the transfer function inversor 405 would be a logarithmic operation.

The transfer function inversor is not only limited to a logarithmic function.

Any 10 mathematical operation that yields a linear energy level error correction signal may be used.

The control signal sensitivity estimator 405 produces an estimate of the slope of the gain versus control signal of the adjustable gain amplifier 105.

The 15 past values of the control signals VCTL, and energy level inverse signal from the transfer function inversor 405 are used to determine future slope estimates.

The output of a control signal sensitivity estimator 407 may be described by the following equation:

SCALE (k) = SCALE '(k-1 )-ALPHA[VCTL(k-1 )-VCTL(k-2)] [EouT(k-l)- 1 ][SCALE (k-1 )] where SCALE'(k) is the present scaled slope estimate signal, ALPHA is a adaptation loop constant, VCTL jS a control signal 25 indexed through time, and EOUT is a scaled average energy level signal indexed through time.

The output of the transfer function inversor 405 is an energy level error correction signal.

The goal of the AGC loop in the preferred embodiment of the present 30 invention is to try to keep the energy error to be zero.

The energy error correction signal is coupled to a subtractor 409.

The subtractor 409 subtracts the energy error correction signal from the present control signal VCTL to generate a new control signal VCTL.

The net output at the 35 subtractor 409 of the gain control signal generator may be 20~19~ 1 2 described by the following equation, for a logarithmic transfer function inversor:

VcTL(k)=vcTL(k~ [scALE (k)/2]LOG1 o[KsEOut(k-1)] where VCTL(k) is the present control signal, SCALE (k) is the present slope estimate signal, Ks is a scaled gain factor and EOUT is a scaled average level signal.

The important elements of the AGC apparatus 1 0 disclosed in this preferred embodiment of the present invention comprise a selective averagef 401, a transfer function inversor 405, a control voltage sensitivity estimator 407 and a subtractor 409.

The selective averager 401 determines an average energy level in a time slot for only a present signal 321.

The averaged energy level value is scaled to a normalized value and coupled to transfer function inversor 405 which uses the normalized average energy value and an estimate of the slope of the gain versus control signal of the adjustable gain amplifier 20 105 to produce an output representing an energy level error correction signal.

The energy level error correction signal may be a positive or negative value.

This value is then subtracted from the present control signal to produce a new control signal which will adjust the adjustable gain 25 amplifier 105 to produce an energy level at the output of the adjustable gain amplifier 105 at a normalized value, in a single cycle of calculation.

When the receive signal encounters conditions that change the level of the energy value, the selective averager 401 determines a new 30 average energy value which is further scaled by the feedback gain adjustor 403 and coupled to the transfer function inversor 405. The transfer function inversor 405 produces an output which has an energy level error signal value which would be subtracted from the now-present 3 5 control signal thereby adjusting the gain of adjustable 20~1982 1 3 gain amplifier 105 to again normalize the energy level at the output of the adjustable gain amplifier 105. A key advantage of the implementation of the preferred embodiment is the speed at which the adjusted normalized energy level is determined.

The accurate knowledge of the gain control transfer function of the adjustable gain amplifier 105, combined with the selective averaging over a span of samples generated only during the effect of the present control signal level 321, allows normalization in only a single cycle of the process by proper calculation of the next control signal, without detrimental effects due to poles and zeros introduced into the feedback ioop by matched filter delay and A/D converter delay.

Now referring to FIG. 5 there is shown a block diagram of the control signal sensitivity estimator 407.

The control voltage sensitivity estimator 407 produces a scaled slope estimate signal.

The scaled slope estimate signal is determined using the change over time of the control signal and the energy level inverse signal.

The change in control signal is determined by the difference between the present control signal (VCTL(k-1)) and the previous control signal (VCTL(k-2)) at the input of the adjustable gain amplifier 105.

The energy level inverse signal is determined by the transfer function inversor 405.

Energy level inverse correction signal from the transfer function inversor 405 is coupled to a gain stage 501 which reduces the input signal by one-half.

The gain stage 501 transforms the input log-energy-referenced signal into an output log-voltage-referenced signal.

The gain stage of one-half being equivalent to taking a square root prior to the logarithmic operation.

The output voltage signal is then coupled to 10X function which compensates the signal for the exponential characteristics of the adjustable gain amplifier 105.

The 10X function 503 3 5 transforms a log-referenced input signal into a linear 20S1982 1 4 referenced output signal. A value of one 505 jS subtracted from the linear- voltage-referenced signal in a summer function 507.

The output of the summer function 507 jS a linear-voltage referenced-estimation of the difference 5 between the present energy level signal and an ideal normalized signal value of one, 505.

A differential amplifier 509 subtracts the present control signal (VCTL(k 1)) from the previous control (VCTL(k-2)) signal.

The previous control signa' is 10 determined by a unit delay function 511.

The change in control signal output from the differerrtial amplifier 509 is coupled to a gain stage 513.

The gain stage 513 multiplies its input by a small number to produce a scaled output.

The small number determines the response time 15 for measuring the change in amplifier gain versus control signal characteristics.

The small number is determined by the noise we wish to filter out of the receive signal, which perturbs the measurement of the response of the gain control process, which response is used to correct the 2 0 estimate of the gain versus control characteristics of the adjustable-gain amplifier.

Without the scaling factor the noise present in a receive signal could cause large and rapid fluctuations of estimate of the gain versus control signal of the adjustable-gain amplifier.

By scaling the 25 change in control voltage by a small number our average estimates are good but each individual measurement could be off by a large amount due to noise.

Therefore the scaling gain amplifier 513 gradually corrects the change in gain versus control voltage estimate without large 30 fluctuations.

The output of the gain amplifier 513 jS coupled to a mixer 515 along with the output of a unit delay function 517.

These inputs are combined to produce an output which is coupled to the mixer 517.

The input from the mixer 515 jS combined with the input from the 35 summer 507 to produce an output for the differential 20S198~ 1 5 amplifier 519.

The differential amplifier 519 subtracts the output from the mixer 517 from the output from the unit delay function 517 to produce an output which is coupled into the unit delay function 517 and also coupled 5 to the gain amplifier 521.

The output of gain amplifier 521 is coupled to the transfer function inversor 405.

The mixers 515 and 517, differential amplifier 519, the unit delay function 517 and the gain stage 521 use the change in energy estimate and the change in control voltage 10 estimate to produce a scaled slope estimate signal as its output.

An advantage of tracking the gain versus control signal transfer function of adjustable gain amplifier 105 is that characteristics specific to each physical amplifier 15 can be taken into account.

Individual characteristics of adjustable gain amplifier 105 of interest include variations in temperature response and manufacturing process tolerances.

20 We claim: